On our latest guest blog for partner S4RB we take some publicly available Tweets looking specifically at packaging. It’s a timely example to use as it is a growing topic as consumers voice more and more environmental concerns, as well as the usual quality issues relating to packaging.
In a generic text analysis model, you might be lucky to pick up packaging issues at all, or at best to pick them up and assign positive or negative sentiment. However, this doesn’t help to action anything, not without further reading and coding. One must also figure out what the themes are to code in the first place.
With ‘human-in-the-loop’ software like PrediCX and the S4RB model specific to grocery retailer, the new signals are referred to a human as they appear so that nothing is missed, nor does it have to be guessed. The data truly speaks for itself!
@Tesco check this out! Bacon I can open one handed. Unlike your packaging that I have to attack with a knife because for years the pull tab has not worked once. @AldiUK #voodoomagic #bacon
Beyond sentiment this richness allows brand owners to understand competitive advantage or disadvantages, which can feed into either marketing or product development. The ease of opening on Aldi’s product will be for more than just bacon!
Also, there’s a hint of long-standing Tesco customer so can add label: “loyal customer”. These tags can both be used to help improve packaging, avoid serious issues, and also improve the brand’s standing to competitors in terms of the features that customers mention. What’s interesting is that a longstanding, loyal Tesco shopper has made an unsolicited comment to Tesco about a competitor. Have they switched? Imploring their favoured brand to improve?
In a large company an internal helpdesk can be as large and complex as any external CRM and keeping down costs whilst keeping up service levels are high priorities.
How PrediCX was applied to ServiceNow data
The customer was an enterprise providing software and consultancy with around 5,000 employees and around 30,000 tickets per year. Warwick Analytics applied its PrediCX software to the helpdesk tickets from ServiceNow (it can also work with BMC Remedy, Zendesk, Salesforce and others). After a short training period of a couple of days, PrediCX was already classifying unstructured data i.e. the text notes within the tickets as well as the notes of the solutions and corrective actions. There were several use cases:
• Early warning of issues and hints for root causes to minimise risk and lost productivity
• Root causes of common issues to obviate tickets and cost, and improve service levels
• Identify opportunities for automation and self-serve both direct and to support agents
• Automated, accurate classification
The view was both retrospective and forward-looking, i.e. to identify the opportunities it could have saved in the past had it been implemented at the time, as well as identifying opportunities going forward.
What PrediCX found
The support teams spend much of their time dealing with these situations rather than enhancing the service. Whilst there will always be emergency situations, the opportunities are to spot the common issues as quickly as possible with alerts, as well as identifying the root causes from the notes of investigated tickets. This can isolate the relevant failure mode quickly and hint at the corrective action required, as well as identifying whether the issue is new or a repeat of something in the past. This is compared to the alternative of manually classification which does not pick out the rich detail of the ticket symptoms (or solutions) and is often inconsistent and inaccurate.
Early warning of payroll issues – The insight allows managers to quickly see when certain failure modes are reappearing e.g. the Expense error FM2 which reoccurred from the first incident on 16 March 2018 and reappeared on 4 May 2018. If PrediCX had been used at the time, it would have helped to implement a permanent fix during the first incident. It also would have shrunk the time of impact of issues by providing the earliest warning of an issue and hints at root causes e.g. the Submission error FM2 from 4 March 2018 to 25 March 2018. PrediCX can help obviate future failure modes and facilitate projects to implement preventative and corrective actions that can be executed ‘offline’ without disrupting service levels.
Hidden laptop issues – the analysis revealed a driver incompatibility issue that took 9 months to resolve. With PrediCX, it’s easy to see that it correlates with a particular Windows error and memory error too. Alert triggers within PrediCX would have picked this up.
Opportunities for automation and deflection – PrediCX looked at tickets over a period of 14.5 months and provided an analysis of whether issues are to do with staff growth and activity, wear & tear (for hardware), a repeating issue which can be deflected (i.e. estimates based on common root causes), a repeated issue which can’t be deflected (i.e. where no common root causes) and issues which appear to be non-repeating. These hint at the potential opportunities for deflection and automation. It shows that 25% of the total tickets analysed
could have been deflected and 29% could be automated to some degree. Given there are about 40,000 tickets per annum and based on a typical cost of solving a ticket, it is estimated that PrediCX identified savings of around a third of the cost of the helpdesk. There may be further opportunities to save on wear and tear too, e.g. by further insight into the supply chain and whether alternative suppliers or processes can prolong the life of assets. There are also opportunities to classify the rest of the tickets automatically and more consistently which leads to more accurate triage and resolution.
Warwick Analytics is able to generate actionable insight and automation at both a strategic and tactical level for helpdesks. It enables helpdesks to optimise their costs whilst maintaining service levels to meet the expectations of their internal customers.
When it comes to the restaurant market, along with the rest of the hospitality market, customer tastes can change and their expectations only grow. Every brand needs to stand for a memorable experience.
Warwick Analytics applied its PrediCX software to publicly available reviews, in particular TripAdvisor reviews for London restaurants. The analysis was centred around use cases that would improve the profitability of restaurants including:
• Understanding the issues which drive churn, loyalty, yield and advocacy
• Operational early warning with granular analysis of issues
• Marketing effectiveness in terms of looking at voucher and campaign feedback
• Compare against the competition, by chain and by location
PrediCX is an automated machine learning platform that quickly and accurately generates models for text, using ‘human-in-the-loop’ technology i.e. it only needs minimum input from a non-data scientist. It took only a few hours to generate meaningful output, no matter how large the dataset, based on concepts instead of keywords and sentiment scoring.
What PrediCX found
By looking at all of the second level concepts being talked about by diners in London, aggregated for all reviewed restaurants and normalised as a proportion of total reviews, we can see the concepts that diners talk about most frequently. The two most common issues are both negative – small portions and bland food – followed by a positive one – good drinks selection etc. This view can be aggregated in any way required: By geography, by branch, over time, segment, sector etc.
Overall, bad service was the main driver of churn at Level1 and at Level2 – small portions, bland food, poor cooking and rudeness were the main causes. This could be used at the brand or branch level to set KPIs and ensure that levels are maintained appropriately.
At Level 1, excellent food and ambience were cited although excellent service was less essential. At Level 2, the view, drinks selection and entertainment were drivers.
Loyalty and churn indicators were also analysed for one specific London restaurant, TGI Friday’s in Covent Garden.
PrediCX can be used to pick up the reviews which contain concepts for churn, negative advocacy or the root causes of churn. They can be quickly intercepted by the restaurant to try to recover the customers with an appropriate message or offer, as well as decreasing the negative advocacy on the web. It can also be used for marketing effectiveness, e.g. picking up concepts of where people have used vouchers and the associated experience and loyalty.
Identify fake reviews
One of the banes of social media is the growing issue of fake, solicited and gamified reviews, the latter being where review sites work with companies to encourage or invite positive reviews and discourage negative reviews in a non-transparent way. There is no way to stop this entirely, but PrediCX can help to train on known fake reviews, remove suspicious or simply glib reviews such as: “everything” [5 stars], or “excellent” [5 stars]. Clearly more reliable data would come from a properly weighted survey, or from the CRM system.
Warwick Analytics is able to generate actionable insight at both a strategic and tactical level for of opportunity for any chain of restaurants, bars or other hospitality. It enables chains to maintain their brand promise whilst at the same time having the ability to react quickly to issues at an aggregate and even a specific customer level to optimise customer experience and maintain loyalty.
A lot has been made of digital transformation, and how many businesses are using self-serve web-based applications to engage with their customers, employees and other stakeholders to be able to enhance and in some cases reinvent the customer experience often with both a stickier customer journey and lower service costs. Uber and AirBnB are often held up as the poster-boys but there are many businesses who are not ‘digital native’ companies emulating in their own way.
As with so many buzzphrases, there is usually a less-sexy way of saying the same thing which has been around for a long-time. In the field of customer interaction, most people will think of digital transformation as the growth in chatbots and social media-enabled communication. However I would argue that the main bastion of change has to be directed at FAQs.
Sexy or not, FAQs used to be the only way to read about self-help and avoid calling a contact center. They are frequently cited as inherently flawed as these blogs from the UK Government and eloquently in this technical writers blog. Yet if you stop and think, a well-structured FAQ if it is searchable with natural language is a critical asset as it is really the same thing as a chatbot, but perhaps without the charm or manners.
In a more measured manner, really FAQs are part of a spectrum of communication channels (one-way and two) where customers can solve problems. They sit alongside forums, social media, chatbots, chat, phone and email (see diagram below).
Also surveys and reviews can trigger interactions depending on the content. The current state of most organisations is that all these elements are separate silos, and whilst the customer experience team are trying hard to break these silos down, there are few who would see FAQs as on the same spectrum as chatbots and forums. Also people’s expectations of FAQs are to see a laundry list of requests which is not how they desire to interact. Imagine though if you could write your query in any way you wanted into a search bar and it would retrieve the correct response. Imagine also that the search was entirely consistent across all channels. Is this just a fantasy?
Machine learning for text is capable to classify interactions to be able to automate responses to natural language. However as we see with chatbot fails, it is hard to get this right due to the complexity and variability of human dialogue and chatbot containment rates are still below where their proprietors would want them to be. Further complexity is that human dialogue varies immensely across the channels: People don’t write in emails how they use chat which is different again to forums, nor how they speak or write a complaint, or even fill in a survey. By way of an example, a study at an airline found that the average topic in a chat was just over one whereas in a call it was nearer to two and in a complaint it was two and a half. People use different channels for different things, and also use different channels for the same thing in different ways. If the company is trying to classify aka tag or ‘label’ each interaction, then it will very easily fall into the trap of having different categories or tags for different channels, not by design but because it is hard to normalise them whatever technology you’re using. This phenomenon doesn’t really have a formal name but it is rife and disruptive. The ideal is some kind of ‘homogenization’ of the tags i.e. so that “late shipping” can be the same concept whatever channel. This then allows the guardians of the customer journey to understand what’s going wrong (and right), get a global view, and also understand if customers are calling back about the same thing on a different channel because they didn’t get it resolved. This also means that the customer journey and knowledge base can be fixed once for each breach, in the knowledge that it is fixing things across the board.
Machine learning can help this harmonization process although it is fraught with challenges, not least because the models for each tag need to be built especially for each channel, for example the “late shipping” tag for chat will need to be a different model to the “late shipping” tag for email or complaint. What data scientists know is that the process of building the machine learning models is intense: New York Times estimated that up to 80% of a data scientist’s time is spent “data wrangling”. CrowdFlower estimates “data preparation” at 80%. Further, assumptions and errors are an inevitable part of the process where human judgement and skill is required. More than this, 76% of data scientists view data preparation as the least enjoyable part of their work. Furthermore someone needs to build a training set for the models and that typically involves a human somewhere labelling the various interactions into topics and a topology that can drive the correct response. This is laborious in a linear fashion.
There are a number of different approaches to this problem. One company addressing this problem in a novel way is Warwick Analytics which is a spin-out from The University of Warwick. It has developed a proprietary technology called ‘Optimized Learning’ which puts a ‘human-in-the-loop’ in a very effective way: What this means is that the technology classifies the customer interactions in a meaningful way but when its certainty is low, it asks for assistance for a human to classify or ‘label’ the interactions which provide the most information back to training the models. Therefore it is theoretically and practically guaranteed to involve the minimum human interaction to maximise the performance of the models and hence the accuracy. The human trainer can be offline, as well as involving the customer in certain circumstances. The company has worked with many enterprises to improve chatbots, automate contact centers, complaints handling and improve the quality of self-service and FAQs.
So in conclusion, FAQs are an old-fashioned and much discredited digital experience, yet in the new world of digital transformation and harmonization, they can come back to center stage thanks to some clever technology and the human-in-the-loop.